Video detection method, apparatus, device, storage medium, and computer program product

Abstract

The application discloses a video detection method and device, equipment, storage medium and computer program product, which can be applied to various fields or scenes such as artificial intelligence, cloud technology and vehicle-mounted video, wherein the method comprises the following steps: obtaining a video feature set of a target video, the video feature set comprising M target video features and a viewing state corresponding to each target video feature, each target video feature being a video feature of one target video segment in M target video segments obtained by video segmentation on the target video; obtaining a to-be-detected recommended video from a video recommendation pool, performing video segmentation on the to-be-detected recommended video to obtain N recommended video segments; determining the video features of each recommended video segment to obtain N recommended video features; and determining a spoiler detection result of the to-be-detected recommended video based on the video feature set of the target video and the N recommended video features. Through the method, the spoiler detection result of the to-be-detected recommended video can be accurately determined.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a video detection method, a video detection device, a computer equipment, a computer-readable storage medium, and a computer program product. Background Technology

[0002] When video playback applications showcase a video (e.g., a TV series, movie, variety show), they often also display recommended short videos related to the video's content. These recommended short videos are uploaded by ordinary or professional users and are not the original content. If the displayed recommended short videos contain content that the user has not yet watched (such videos are called spoilers), and the user chooses to watch these recommended videos beforehand, they will know the original video content. This usually reduces the user's interest in watching the original video, and they may not choose to watch it at all, thus seriously affecting the original video's rights, such as viewership. Therefore, it is essential to know how to block spoiler videos. A prerequisite for blocking spoiler videos is accurately identifying them, but how to accurately determine whether a recommended short video is a spoiler is currently an unsolved problem. Summary of the Invention

[0003] This application provides video detection methods, apparatus, devices, storage media, and computer program products, which can accurately determine the spoiler detection results of recommended videos to be detected.

[0004] One embodiment of this application discloses a video detection method, the method comprising:

[0005] Obtain a video feature set of the target video; wherein, the video feature set includes M target video features and the viewing state corresponding to each target video feature, each target video feature is a video feature of one of the M target video segments, the M target video segments are obtained by video segmentation of the target video; different target video features correspond to different target video segments, and the viewing state corresponding to a target video feature is the viewing state of the target video segment corresponding to the target video feature; M is a positive integer greater than 1;

[0006] Obtain recommended videos to be detected from the video recommendation pool, and perform video segmentation on the recommended videos to be detected to obtain N recommended video segments, where N is a positive integer greater than 1;

[0007] Determine the video features of each of the recommended video segments to obtain N recommended video features;

[0008] Based on the video feature set of the target video and the features of the N recommended videos, the spoiler detection result of the recommended video to be detected is determined.

[0009] One embodiment of this application discloses a video detection device, which includes:

[0010] An acquisition unit is used to acquire a video feature set of a target video; wherein, the video feature set includes M target video features and the viewing state corresponding to each target video feature, each target video feature is a video feature of one of the M target video segments, the M target video segments are obtained by video segmentation of the target video; different target video features correspond to different target video segments, and the viewing state corresponding to a target video feature is the viewing state of the target video segment corresponding to the target video feature; M is a positive integer greater than 1;

[0011] The processing unit is used to obtain the recommended video to be detected from the video recommendation pool, and to perform video segmentation on the recommended video to be detected to obtain N recommended video segments, where N is a positive integer greater than 1;

[0012] A determining unit is used to determine the video features of each of the recommended video segments to obtain N recommended video features; based on the video feature set of the target video and the N recommended video features, the spoiler detection result of the recommended video to be detected is determined.

[0013] One aspect of this application discloses a computer device, which includes a processor adapted to implement one or more computer programs; and a computer storage medium storing one or more computer programs, the one or more computer programs being adapted to be loaded by the processor and executed in accordance with the video detection method described above.

[0014] One aspect of this application discloses a computer-readable storage medium storing one or more computer programs adapted to be loaded by a processor and executed by the video detection method described above.

[0015] One embodiment of this application discloses a computer program product, which includes a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the aforementioned video detection method.

[0016] In this embodiment, a video feature set of the target video is obtained. The video feature set includes M target video features and the viewing state corresponding to each target video feature. Each target video feature is a video feature of one of the M target video segments obtained by video segmentation of the target video. The viewing state corresponding to the target video feature is the viewing state of the target video segment corresponding to the target video feature. A recommended video to be detected is obtained from the video recommendation pool. The recommended video to be detected is segmented to obtain N recommended video segments. Then, the video features of each recommended video segment are determined to obtain N recommended video features. Finally, based on the video feature set of the target video and the N recommended video features, the spoiler detection result of the recommended video to be detected is determined. This method determines the spoiler detection result of the recommended video based on the video segment features and viewing status of each video segment of the target video and the video segment features of each video segment of the recommended video to be detected. On the one hand, spoiler detection can be completed based only on the video features (i.e., the features of video screen information). Compared with the method of combining video screen information, video subtitle information, and video bullet screen information for spoiler detection, the processing method is simpler and the amount of data is smaller, so spoiler detection can be completed quickly. On the other hand, dividing the video into multiple video segments and performing spoiler detection based on the video segment features of each video segment can achieve fine-grained feature comparison detection, thereby effectively improving the accuracy of spoiler detection results. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a network architecture diagram of a video detection system disclosed in an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating a video detection method disclosed in an embodiment of this application;

[0020] Figure 3 This is a network structure diagram of a feature extraction module disclosed in an embodiment of this application;

[0021] Figure 4 This is a video recommendation page corresponding to a target video, as disclosed in an embodiment of this application;

[0022] Figure 5 This is a flowchart illustrating another video detection method disclosed in an embodiment of this application;

[0023] Figure 6 This is a flowchart illustrating another video detection method disclosed in the embodiments of this application;

[0024] Figure 7 This is a schematic diagram of the structure of a video detection device disclosed in an embodiment of this application;

[0025] Figure 8 This is a schematic diagram of the structure of a computer device disclosed in an embodiment of this application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] To effectively improve the user's viewing experience, this application proposes a video detection method that can accurately determine the spoiler detection result of a recommended video to be detected. That is, it can accurately determine whether the recommended video to be detected is a spoiler video of the target video or a non-spoiler video of the target video. Based on this, when showing users recommended videos associated with the target video, spoiler videos of the target video can be accurately filtered out, and only non-spoiler videos of the target video can be shown to the user. This can effectively avoid the problem that users will lose interest in watching the original work due to watching spoiler videos and will not choose to watch the original work, thereby protecting the rights of the original work of the target video, such as the number of views.

[0028] The video detection method provided in this application embodiment can be implemented based on AI (Artificial Intelligence) technology. AI refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain optimal results. AI technology is a comprehensive discipline involving a wide range of fields; the video detection method provided in this application embodiment mainly involves machine learning (ML) technology within AI. Machine learning typically includes techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0029] In feasible embodiments, the video detection method provided in this application can also be implemented based on cloud technology and / or blockchain technology. Specifically, it may involve one or more of cloud technologies, such as cloud storage, cloud database, and big data. For example, data required to execute the video detection method (e.g., video feature set of the target video, recommended videos to be detected, etc.) can be obtained from a cloud database. As another example, the data required to execute the video detection method can be stored on the blockchain in the form of blocks; the data generated by executing the video detection method (e.g., recommended video features, spoiler detection results, etc.) can be stored on the blockchain in the form of blocks; furthermore, the data processing device executing the video detection method can be a node device in a blockchain network.

[0030] The video detection method provided in this application embodiment can be applied to... Figure 1 For the network architecture shown, please refer to [link / reference]. Figure 1 This is a network architecture diagram of a video detection system disclosed in an embodiment of this application. The video detection system 100 may include at least a terminal device 101 and a computer device 102. The terminal device 101 and the computer device 102 can be connected to each other, and the connection method may include wired connection and wireless connection, which is not limited here. In the specific implementation process, the terminal device 101 is mainly used to display results. In this application, it can display the target video and non-spoiler recommended videos related to the target video. The computer device 102 is mainly used to obtain the video feature set of the target video and the recommended videos to be detected, then process the recommended videos to obtain the features of the recommended videos, and then determine the spoiler detection result of the recommended videos to be detected based on the video feature set of the target video and the features of the recommended videos.

[0031] In one possible implementation, the terminal device 101 mentioned above includes, but is not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc.; the computer device 102 mentioned above is a data processing device, which can be a server. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Figure 1 This is merely an illustrative network architecture diagram of a video detection system and is not intended to be limiting. For example, Figure 1The computer device 102 can be deployed as a node in the blockchain network, or the computer device 102 can be connected to the blockchain network, so that the computer device 102 can upload the video feature set of the target video, the recommended videos to be detected in the video recommendation pool, and the spoiler detection results to the blockchain network for storage, so as to prevent the internal data from being tampered with, thereby ensuring data security.

[0032] Based on the aforementioned video detection system, the video detection method of this application embodiment can generally include: acquiring a video feature set of a target video; wherein, the video feature set includes M target video features and the viewing state corresponding to each target video feature, each target video feature is a video feature of one of the M target video segments, the M target video segments are obtained by video segmentation of the target video; different target video features correspond to different target video segments, and the viewing state corresponding to a target video feature is the viewing state of the target video segment corresponding to the target video feature; M is a positive integer greater than 1; acquiring a recommended video to be detected from a video recommendation pool, segmenting the recommended video to be detected to obtain N recommended video segments, N being a positive integer greater than 1; determining the video features of each of the recommended video segments to obtain N recommended video features; and determining the spoiler detection result of the recommended video to be detected based on the video feature set of the target video and the N recommended video features.

[0033] Through the embodiments of this application, the target video features of the target video and the recommendation features of the recommended video to be detected can be compared. By combining the viewing status of the video segments corresponding to each target video feature, it is possible to accurately determine whether the recommended video to be detected is a spoiler video. Furthermore, spoiler videos can be hidden, and only non-spoiler recommended videos and target videos can be shown to users. This can improve the user's viewing experience and, to a certain extent, increase the number of users watching the target video.

[0034] It should be noted that in the specific embodiments of this application, the data involved, such as the target video and the recommended videos to be detected in the recommendation pool, are all authorized by the user. When the above embodiments of this application are applied to specific products or technologies, the data used must be authorized or agreed to by the user, and the collection, use, and processing of the relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0035] Please see Figure 2 This is a flowchart illustrating a video detection method disclosed in an embodiment of this application. The video detection method described in this embodiment can be performed by... Figure 1 The computer device shown executes the video detection method, which includes, but is not limited to, the following steps:

[0036] S201: Obtain the video feature set of the target video.

[0037] In feasible embodiments, the target video can be any of the following: TV series, movies, variety shows, or animations. The video feature set of the target video includes M target video features and the viewing state corresponding to each target video feature. Each target video feature is a video feature of one of the M target video segments. That is, one target video feature among the M target video features is a video feature of one of the M target video segments. During feature extraction, the M target video segments are obtained by video segmentation of the target video. The viewing state of each target video feature is the same as the viewing state of the target video segment corresponding to that feature. The viewing state includes a watched state and an unwatched state. The viewing state of the target video is acquired simultaneously with the target video. That is, when the target video is in a watched state, the corresponding target video feature is also in a watched state; when the target video is in an unwatched state, the corresponding target video feature is also in an unwatched state. The method of acquiring the viewing state is relatively flexible and can be acquired in real time based on the user's viewing status.

[0038] S202: Obtain the recommended video to be detected from the video recommendation pool, and perform video segmentation on the recommended video to be detected to obtain N recommended video segments, where N is a positive integer greater than 1.

[0039] After identifying the target video, a series of recommended videos to be detected can be obtained from the video recommendation pool of the video application. Then, the recommended videos to be detected are segmented to obtain N recommended video segments, where N is a positive integer greater than 1.

[0040] In feasible embodiments, for any recommended video to be detected, a video segmentation module can be used to segment the video during video segmentation. This module segments the video based on the video segmentation information of the recommended video to be detected. The video segmentation information includes one or both of black frame information and scene transition frame information. Based on one or both of the black frame information and scene transition frame information, the recommended video to be detected can be segmented to obtain N recommended video segments. That is, the recommended video to be detected can be segmented based on either black frame information or scene transition frame information (one of which is a video segmentation point), or it can be segmented based on both black frame information and scene transition frame information (both of which are video segmentation points).

[0041] Black frame information can be determined based on the pixel values ​​of video frame images. The specific determination process can include: first, extracting each frame image from the recommended video to be detected; then calculating the average RGB pixel value (0 to 255) of each frame image and comparing it with a set threshold. Values ​​less than the set threshold are considered black frame information. Black frames are typically used for transitions in videos, so detected black frames can usually be used as transition points in video segments. Scene switching frame information can be determined based on the differences between video frame images. The specific determination process can include: extracting each frame image from the recommended video to be detected and converting each frame image from the RGB color space to the HSV color space; then calculating the changes between adjacent frames in the HSV color space, that is, calculating the changes between hue (H), saturation (S), and brightness (V). If the changes are greater than a set threshold, a scene switch is considered to have occurred, and the corresponding video frame is determined as scene switching frame information.

[0042] In this application, the recommended videos to be detected may or may not be associated with the target video. Since this application aims to filter spoiler videos for the target video, the recommended videos to be detected can be considered to be associated with the target video. These recommended videos are typically content uploaded by ordinary or professional users, i.e., user-generated content (UGC). Optionally, the recommended video to be detected may be a video composed of clips from a single episode of a TV series, or a video composed of multiple clips edited from multiple episodes of a TV series; it may also be a video composed of different clips from different TV series. This is not limited to any specific type of video.

[0043] S203: Determine the video features of each recommended video segment to obtain N recommended video features.

[0044] In one possible implementation, a video feature extraction module can be used to extract features for each recommended video segment. The network structure diagram of the feature extraction module can be found in [reference needed]. Figure 3 The system consists of a 3D patch partitioning module and four stages. Each stage comprises two parts: Stage 1 includes a linear embedding module and a video processing module (Swin Transformer Block); Stages 2 through 4 each include a patch merging module and a video processing module. The 3D patch partitioning module divides the system into equal patches; the linear embedding module performs linear mapping on these patches; and the patch merging module is similar to pooling, but unlike pooling, which loses information, the patch merging module does not. Figure 3For example, the input is a recommended video clip of size T×H×W×3, where T can be set to 32, indicating that 32 frames are sampled from all frames of the recommended video clip. The sampling method can be chosen freely, and different tasks may have different sampling methods. Generally, equal-interval sampling can be used. After the input is processed by the 3D block partitioning module, it becomes a vector of (T / 2)×(H / 4)×(W / 4)×96. Since the patch size is (2,4,4), which represents the dimensions of time, height, and width, 96 is used because 2×4×4×3=96, which is the value of the three RGB channels (R for Red, G for Green, and B for Blue) of all pixels within a patch. The 3D block partitioning module is followed by a linear mapping module. If the dimension of the linear mapping module is 96, then the size after the linear mapping module is 96. Then, it will be processed by multiple video processing modules and block merging modules to obtain the recommended video features of the recommended video segment. The whole process imitates the downsampling process in the CNN module, which allows the model to generate features for different scales.

[0045] S204: Based on the video feature set of the target video and the features of N recommended videos, determine the spoiler detection result of the recommended video to be detected.

[0046] In this embodiment, the spoiler detection result is either that the recommended video to be detected is a spoiler video or that it is a non-spoiler video. A spoiler video is video content that the user has not yet watched, and a non-spoiler video is video content that the user has already watched. Based on the above description, the main approach is to search for N recommended video features in the video feature set of the target video, and then determine the spoiler detection result of the recommended video to be detected based on the search results.

[0047] In one possible implementation, the spoiler detection result of the recommended video to be detected, based on the video feature set of the target video and N recommended video features, can be determined as follows: the N recommended video features are compared with each of the target video features in the video feature set of the target video to obtain the comparison result. If the comparison result determines that there is a first type of recommended video feature among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined as the first spoiler detection result. If the comparison result determines that there is no first type of recommended video feature among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined as the second spoiler detection result. The first spoiler detection result indicates that the recommended video to be detected is a spoiler video of the target video, and the second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video.

[0048] In this scenario, the first type of recommended video feature has a matching target video feature in the video feature set, and the viewing state corresponding to the target video feature matching the first type of recommended video feature in the video feature set is "not watched". In other words, if among the recommended video features of the video to be detected, there is a recommended video feature that can find a matching target video feature in the video feature set, and the viewing state corresponding to that target video feature is "not watched", then it indicates that the recommended video to be detected is a spoiler video.

[0049] In one embodiment, if, based on the comparison results, K out of N recommended video features are determined to be second-type recommended video features, and NK out of the remaining K recommended video features are determined to be third-type recommended video features, then it can be determined that no first-type recommended video features exist among the N recommended video features. Here, K is an integer greater than or equal to 1 and less than or equal to N. The second-type recommended video features have matching target video features in the video feature set, and the viewing state corresponding to the target video features matching the second-type recommended video features in the video feature set includes the "watched" state; the third-type recommended video features do not have matching target video features in the video feature set.

[0050] For example, a recommended video to be detected, segmented into 5 recommended video clips, corresponds to 5 recommended video features. Two of these features belong to the second category, meaning there are two matching video features in the feature set, and the corresponding video clips are in a "watched" state. Meanwhile, three features belong to the third category, meaning there are no matching video features in the feature set. Therefore, the recommended video clips corresponding to these three features are unrelated to the target video; they could be clips edited from other videos (unrelated to the target video) or content filmed by the uploader that is unrelated to the target video. Since the two recommended video clips related to the target video are both in a "watched" state, the recommended video to be detected is a non-spoiler video.

[0051] When determining whether recommended video features exist in the video feature set, the nearest neighbor retrieval method can be used. This method primarily calculates the similarity between features to determine whether two video segments belong to the same video segment. The similarity calculation mainly involves calculating the distance between features, such as Euclidean distance or cosine distance. Specifically, it calculates the Euclidean or cosine distance between the recommended video features and the target video features in the video feature set. When the calculated value reaches a preset condition, such as a set threshold (e.g., 0.9), the corresponding target video feature is considered a match for the recommended video feature.

[0052] After determining the spoiler detection results for the recommended videos to be detected, the relevant recommended videos can be processed. Spoiler videos are filtered out, while non-spoiler videos are displayed. Specifically, if the spoiler detection result of the recommended video indicates that it is a spoiler for the target video, then the recommended video is placed in the spoiler video pool of the target video; if the spoiler detection result indicates that it is a non-spoiler video of the target video, then it is placed in the non-spoiler video pool of the target video. Further, the target recommended video is obtained from the non-spoiler video pool of the target video; the relevant information of the target recommended video is sent to the client so that the client can display the relevant information of the target recommended video on the video recommendation page corresponding to the target video in the video playback application. In other words, when playing the target video in the video playback application, the target recommended video can first be determined from the recommended videos placed in the non-spoiler video pool, and then the relevant information of the target recommended video can be displayed on the video recommendation page. As users watch videos, over time, when a spoiler video in the spoiler video pool is viewed, it can be moved from the spoiler video pool to the non-spoiler video pool, and related information can be displayed on the video recommendation page. The information displayed for the recommended video can take various forms, including an icon link or an icon showing the video's cover image.

[0053] Optionally, if the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a spoiler video of the target video, then a spoiler label is added to the recommended video to be detected; if the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a non-spoiler video of the target video, then a non-spoiler label is added to the recommended video to be detected. When the target video is played in the video playback application, the corresponding target recommended video can be obtained based on the non-spoiler label, and the relevant information of the target recommended video can be displayed on the video recommendation page. As the user watches, the viewing status of some spoiler videos will change. When a user watches a spoiler video, the spoiler label of that spoiler video can be converted into a non-spoiler label.

[0054] The video recommendation page corresponding to the target video can be as follows: Figure 4 As shown, 401 is the playback area for the target video, and 402 is the video recommendation display area. 402 can include multiple recommended videos: Recommended Video 1, Recommended Video 2, and Recommended Video 3. All recommended videos displayed in 402 are non-spoiler-free. For example, if episode K of a TV series is playing in 401, then most of the recommended videos in 402 will be video clips related to episodes 1 through K-1. Optionally, the number of recommended videos uploaded by the user may be large, but the video recommendation display area is limited. Therefore, the recommended videos can be filtered to display target recommended videos that meet the criteria.

[0055] This process involves filtering recommended videos based on their playback time to determine the target recommended videos. First, the current playback time of the target video is determined. Then, using this current time as a boundary, the recommended videos closest to the current time are identified. Finally, based on the layout of the video recommendation display area on the video recommendation page, the number of target recommended videos is determined from the remaining recommended videos. If the display area can show 5 videos, then 5 are selected. For example, if the currently playing video is episode K of a TV series, then the recommended videos could be video clips related to episodes 1 through K-1. Based on playback time, video clips from episode K-1 are selected first, followed by episode K-2, and so on, until the required number of target recommended videos is determined.

[0056] This application's embodiments determine the spoiler detection results of the recommended videos based on the video segment features and viewing status of each video segment of the target video and the video segment features of each video segment of the recommended videos to be detected. On the one hand, spoiler detection can be completed solely based on the video features (i.e., the features of video screen information). Compared to methods that combine video screen information, video subtitle information, and video bullet screen information for spoiler detection, the processing method is simpler and requires less data, thus enabling rapid spoiler detection. On the other hand, dividing the video into multiple video segments and performing spoiler detection based on the video segment features of each video segment allows for fine-grained feature comparison detection, thereby effectively improving the accuracy of the spoiler detection results. Based on the accurate spoiler detection results of the recommended videos to be detected, when displaying recommended videos associated with the target video to the user, spoiler videos of the target video can be accurately filtered out, and only non-spoiler videos of the target video can be displayed to the user. This effectively avoids the problem that users' interest in watching the original work is reduced due to watching spoiler videos, thus protecting the rights of the original target video, such as viewership.

[0057] Based on the above description, the spoiler detection result of the recommended video to be detected is determined directly based on the video feature set of the target video and N recommended video features. In the following embodiment, the video feature set can be divided into a subset of viewed video features and a subset of unviewed video features, and then the spoiler detection result of the recommended video to be detected is determined based on the subset of viewed video features, the subset of unviewed video features, and the N recommended video features. See also Figure 5 This is another video detection method disclosed in the embodiments of this application, which may include, but is not limited to, the following steps:

[0058] S501: Obtain the target video, segment the target video to obtain M target video segments.

[0059] For example, when the target video is a TV series, the TV series can be sorted by episode number, and then each episode can be segmented to obtain multiple target video segments for each episode. Then, the multiple target video segments for each episode can be combined together to obtain M target video segments. As another example, when the target video is a movie, the movie can be treated as a whole and segmented to obtain M target video segments.

[0060] As explained above, the video segmentation module can also be used to segment the target video. During segmentation, the video segmentation information is first determined from the target video, and then the target video is segmented based on the video segmentation information to obtain M target video segments. The process of confirming the black frame information and the scene segmentation frame information has been described in detail in step S202 above, and will not be repeated here.

[0061] S502: Determine the video features of each of the M target video segments to obtain the M target video features.

[0062] A video feature extraction module can be used to extract features from M target video segments, resulting in M ​​target video features, with one target video segment corresponding to one target video feature. For details on the video feature extraction module, please refer to [link to details]. Figure 3 The video feature extraction process has been explained in detail above, and will not be repeated here.

[0063] S503: Determine the viewing status of each target video segment among the M target video segments. Based on the viewing status of each target video segment, the correspondence between the M target video segments and the M target video features, determine the viewing status corresponding to each target video feature.

[0064] In feasible embodiments, when segmenting a target video, the viewing status of each target video segment can be determined based on the viewing status of the target video. For example, for a TV series containing N episodes, if the first K episodes have been watched by the user, then the viewing status of the target video segments segmented based on these K episodes is also "watched," and correspondingly, the target video features corresponding to these target video segments are also "watched." Conversely, if episodes K+1 to N are not watched, then the viewing status of the target video segments segmented based on episodes K+1 to N is also "not watched," and correspondingly, the target video features corresponding to these target video segments are also "not watched." For a movie, based on the user's viewing history and time division, the TV series can be divided into two parts: a watched part and an unwatched part. Based on this, the watched part is segmented to obtain multiple target video segments, and the viewing status of the target video features corresponding to these target video segments is "watched." Similarly, the unwatched part is segmented to obtain multiple target video segments, and the viewing status of the target video features corresponding to these target video segments is "not watched."

[0065] Since there is a one-to-one correspondence between the M target video segments and the M target video features, the viewing status of the target video features can be determined based on the viewing status of the target video segments. For example, if the target video segments are divided into multiple target video segments, and two of the target video segments have a "watched" status, then the viewing status of the target video features corresponding to these two target video segments is also "watched".

[0066] S504: Generate a subset of viewed video features based on the target video features corresponding to the viewed state, generate a subset of unviewed video features based on the target video features corresponding to the unviewed state, and generate a video feature set of the target video based on the subset of viewed video features and the subset of unviewed video features.

[0067] In one embodiment, each target video feature whose viewing state is "watched" is placed in the same subset among the M target video features to obtain the watched video feature subset; each target video feature whose viewing state is "not watched" is placed in the same subset among the M target video features to obtain the unwatched video feature subset; then the watched video feature subset and the unwatched video feature subset are placed in the same set to obtain the video feature set of the target video.

[0068] S505: Obtain the recommended video to be detected from the video recommendation pool, perform video segmentation on the recommended video to be detected, and obtain N recommended video segments, where N is a positive integer greater than 1.

[0069] S506: Determine the video features of each recommended video segment to obtain N recommended video features.

[0070] For a detailed description of steps S505 and S506, please refer to steps S202 and S203, which will not be repeated here.

[0071] S507: Compare the N recommended video features with each target video feature in the subset of watched video features, and compare the N recommended video features with each target video feature in the subset of unwatched video features to obtain the comparison results.

[0072] For example, if N is 3, then these 3 recommended video features are compared with each target video feature in the subset of watched video features, and simultaneously these 3 recommended video features are compared with each target video feature in the subset of unwatched video features, to obtain the comparison results.

[0073] S508: If, based on the comparison results, it is determined that there is a first type of recommended video feature among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined as the first spoiler detection result.

[0074] The first spoiler detection result indicates that the recommended video to be detected is a spoiler video of the target video. The first type of recommended video features have a matching target video feature in the unwatched video feature subset, but no matching target video feature in the watched video feature subset. In other words, the first type of recommended video features can be matched in the unwatched video feature subset, but cannot be matched in the watched video feature subset. For example, for a recommended video to be detected with 3 recommended video features, if one of its video features can be matched in the unwatched video feature subset but not in the watched video feature subset, then the recommended video to be detected is determined to be a spoiler video.

[0075] It's important to note that the requirement of meeting the first type of recommended video features before determining if a recommended video is a spoiler video is to exclude special cases that could lead to inaccurate judgments. For example, if someone has watched the first episode of a TV series but not the second, and the recommended video is a recollection of a scene from the first episode in the second episode, then the recommended video features would match both the video features of the first and second episode segments. In this case, it's impossible to determine whether the recommended video is a spoiler video. Therefore, the recommended video features must satisfy the condition that a matching target video feature exists in the subset of unwatched videos, but no matching target video feature exists in the subset of watched videos, before the recommended video is identified as a spoiler video.

[0076] S509: If, based on the comparison results, it is determined that there is no first-class recommended video feature among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined as the second spoiler detection result.

[0077] If, based on the comparison results, K of the N recommended video features are determined to be second-type recommended video features, and NK of the remaining N recommended video features are third-type recommended video features, then it is determined that the first-type recommended video features do not exist among the N recommended video features.

[0078] The second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video. K is an integer greater than or equal to 1 and less than or equal to N, indicating that the recommended videos to be detected corresponding to these N recommended video features are associated with the target video. The second type of recommended video features have a matching target video feature in the already viewed video feature subset; the third type of recommended video features do not have a matching target video feature in the already viewed video feature subset, nor do they have a matching target video feature in the unviewed video feature subset. For example, a recommended video to be detected is divided into 5 video segments, corresponding to 5 recommended video features. Among them, 2 recommended video features belong to the second type of recommended video features, that is, these 2 recommended video features match the target video feature in the already viewed video feature subset, and 3 recommended video features belong to the third type of recommended video features, that is, these 3 recommended video features do not match the target video feature in the already viewed video feature subset, nor do they match the target video feature in the unviewed video feature subset. Then, the recommended video to be detected is a non-spoiler video. Generally, this type of video is a mashup, which combines video clips cut from multiple TV series to create a single video.

[0079] Optionally, based on the comparison results, if each of the N recommended video features belongs to the fourth category of recommended video features, and the fourth category of recommended video features does not match the target video features in the subset of watched video features or the target video features in the subset of unwatched video features, then it indicates that the recommended video to be detected corresponding to these N recommended video features is not related to the target video, and obviously the recommended video to be detected is not a spoiler video of the target video.

[0080] Optionally, since user behavior is dynamic, the number of video episodes a user has watched varies at different times. Therefore, for each recommended video to be detected (a spoiler video), after the recommended video to be detected is uploaded to the video playback application, it can be searched once to first determine the specific position of the recommended video to be detected in the target video, and then, based on the user's viewing progress, determine the spoiler detection result of the recommended video to be detected. In this embodiment of the application, the playback time corresponding to each first type of recommended video feature is first determined. The playback time is the segment playback time of the target video segment corresponding to the target video feature that matches the first type of recommended video feature in the unwatched video feature subset. The playback time of a video clip can be a time interval. For example, if a movie's length is 0:00:00 to 1:40:50 (from 0 seconds to 1 hour, 40 minutes, and 50 seconds), the playback time could be "0:25:34 to 0:32:52". For a TV series, the playback time can be defined as a specific episode number (e.g., which episode), such as episode 4 at the 20th minute, rather than simply a time interval of 20 minutes. Correspondingly, the playback time corresponding to a video clip feature can also be described by episode number. The system retrieves new viewing time information for the target video, including the playback time of the already viewed portion. The new viewing time can be used to indicate the user's progress in watching the target video. This new viewing time is a point in time based on the target video's viewing time. For example, if a movie's length is 0:00:00 to 1:40:50, the new viewing time could be 1:02:56. The movie content before 1:02:56 is what the user has already watched, and the movie content after 1:02:56 is what the user is currently watching. If the playback time of the already viewed portion of the video includes the playback time corresponding to each of the first category of recommended video features, then the spoiler detection result of the recommended video to be detected will be adjusted from the first spoiler detection result to the second spoiler detection result. The second spoiler detection result indicates that the recommended video to be detected is not a spoiler video of the target video.

[0081] For example, when a video playback application detects a new recommended video A, it searches for recommended video A within the complete source of the target video to determine its position within the target video (e.g., which episode in a TV series, or which moment in a movie), and then marks it. If, for a TV series watched by a user, recommended video A corresponds to episode k (the segment's playback time), as the user watches the series sequentially, the number of episodes watched (the new viewing time) is obtained in real-time. Before the user has watched episode k, episode k is considered unwatched, therefore the recommended video is considered a spoiler and is hidden. However, if the user has already watched episode k, episode k is considered watched, therefore the recommended video is considered non-spoiler-free and can be shown to the user.

[0082] Based on the above explanation, let's take a TV series x with N episodes as an example. Figure 6 This also illustrates a video detection method. Based on the user's viewing history, TV series x can be divided into watched videos (from episode 1 to episode K) and unwatched videos (from episode K+1 to episode N). A video segmentation module is used to segment the watched and unwatched videos, resulting in multiple video segments for each episode. Then, a feature extraction module is used to extract features from each video segment, obtaining video features. The video features obtained from watched videos are used as the first feature information set, and the video features obtained from unwatched videos are used as the second feature information set. Simultaneously, recommended videos associated with TV series x are obtained. Multiple recommended videos can be acquired. The video segmentation module is then used to segment the recommended videos, resulting in multiple video segments. Again, the feature extraction module extracts features from each video segment, obtaining video features. The video features obtained from the recommended videos are used as the reference feature information set. Finally, each feature in the reference feature information set is subjected to a nearest neighbor search in the first and second feature information sets to determine the spoiler detection result for each recommended video. During the search, there are usually two results: if the video can be found in the first feature information set but not in the second feature information set, it means that the recommended video contains content that the user has already watched and is a non-spoiler video, which can be shown to the user; if the video cannot be found in the first feature information set but can be found in the second feature information set, it means that the recommended video contains content that the user has not yet watched and is a spoiler video, which needs to be filtered and cannot be shown to the user.

[0083] In this embodiment, we divide the video feature set into a subset of watched video features and a subset of unwatched video features, and then determine the spoiler detection result of the recommended video to be detected based on the subset of watched video features, the subset of unwatched video features, and N recommended video features. In this way, we can accurately and quickly determine the spoiler videos and non-spoiler videos in the video recommendation pool, thereby improving the user's viewing experience.

[0084] Based on the above method embodiments, this application also provides a schematic diagram of the structure of a video detection device. See also... Figure 7 This is a schematic diagram of the structure of a video detection device provided in an embodiment of this application, corresponding to the computer equipment described above. Figure 7 The video detection device shown can operate the following units:

[0085] The acquisition unit 701 is used to acquire a video feature set of a target video; wherein, the video feature set includes M target video features and the viewing state corresponding to each target video feature, each target video feature is a video feature of one of the M target video segments, the M target video segments are obtained by video segmentation of the target video; different target video features correspond to different target video segments, and the viewing state corresponding to a target video feature is the viewing state of the target video segment corresponding to the target video feature; M is a positive integer greater than 1;

[0086] The processing unit 702 is used to obtain a recommended video to be detected from the video recommendation pool, and to perform video segmentation on the recommended video to be detected to obtain N recommended video segments, where N is a positive integer greater than 1;

[0087] The determining unit 703 is used to determine the video features of each of the recommended video segments to obtain N recommended video features; based on the video feature set of the target video and the N recommended video features, the spoiler detection result of the recommended video to be detected is determined.

[0088] In one possible implementation, the acquisition unit 701 is further configured to acquire the target video, perform video segmentation on the target video, and obtain M target video segments;

[0089] The determining unit 703 is further configured to determine the video features of each of the target video segments to obtain M target video features; determine the viewing state of each of the target video segments, and based on the viewing state of each of the target video segments, the correspondence between the M target video segments and the M target video features, determine the viewing state corresponding to each of the target video features; generate a subset of viewed video features based on the target video features whose corresponding viewing state is "viewed", generate a subset of unviewed video features based on the target video features whose corresponding viewing state is "not viewed", and generate a video feature set of the target video based on the subset of viewed video features and the subset of unviewed video features.

[0090] In one possible implementation, when the determining unit 703 determines the spoiler detection result of the recommended video to be detected based on the video feature set of the target video and the N recommended video features, it is specifically used for:

[0091] The N recommended video features are compared with each target video feature in the subset of watched video features, and the N recommended video features are compared with each target video feature in the subset of unwatched video features to obtain comparison results.

[0092] If, based on the comparison results, it is determined that a first type of recommended video feature exists among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined as the first spoiler detection result.

[0093] If, based on the comparison results, it is determined that the first type of recommended video feature does not exist among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined to be the second spoiler detection result.

[0094] Wherein, the first spoiler detection result indicates that the recommended video to be detected is a spoiler video of the target video, and the second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video; the first type of recommended video feature has a matching target video feature in the unwatched video feature subset, but no matching target video feature in the watched video feature subset.

[0095] In one possible implementation, the determining unit 703 is further configured to:

[0096] If, based on the comparison results, K of the N recommended video features are determined to be second-type recommended video features, and NK of the remaining N recommended video features are third-type recommended video features, then it is determined that the first-type recommended video features do not exist among the N recommended video features.

[0097] Wherein, K is an integer greater than or equal to 1 and less than or equal to N; the second type of recommended video feature has a matching target video feature in the subset of watched video features; the third type of recommended video feature does not have a matching target video feature in the subset of watched video features, nor does it have a matching target video feature in the subset of unwatched video features.

[0098] In one possible implementation, the determining unit 703 is further configured to determine the playback time corresponding to each of the first type of recommended video features, wherein the playback time is the segment playback time of the target video segment corresponding to the target video feature that matches the first type of recommended video feature in the subset of unwatched video features;

[0099] The acquisition unit 701 is further configured to acquire new viewing time information of the target video, the new viewing time information including the video playback time of the viewed portion of the target video;

[0100] The processing unit 702 is further configured to adjust the spoiler detection result of the recommended video to be detected from the first spoiler detection result to the second spoiler detection result if the playback time of the already viewed portion of the video includes the playback time corresponding to each of the first type of recommended video features. The second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video.

[0101] In one possible implementation, when the processing unit 702 performs video segmentation on the recommended video to be detected to obtain N recommended video segments, it is specifically used for:

[0102] Video segmentation information is determined from the recommended videos to be detected. The video segmentation information includes one or both of black frame information and scene switching frame information.

[0103] Based on the video segmentation information, the recommended video to be detected is segmented to obtain N recommended video segments.

[0104] In one possible implementation, the processing unit 702 is further configured to, if the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a spoiler video of the target video, then put the recommended video to be detected into the spoiler video pool of the target video;

[0105] If the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a non-spoiler video of the target video, then the recommended video to be detected is placed into the non-spoiler video pool of the target video.

[0106] In one possible implementation, the acquisition unit 701 is further configured to acquire a target recommended video from the non-spoiler video pool of the target video; and send the relevant information of the target recommended video to the client, so that the client displays the relevant information of the target-detected recommended video on the video recommendation page corresponding to the target video in the video playback application.

[0107] It is understood that the functions of each functional unit of the video detection device provided in this application embodiment can be specifically implemented according to the methods in the above method embodiments, and the specific implementation process can be referred to the relevant descriptions in the above method embodiments, which will not be repeated here.

[0108] In feasible embodiments, the video detection device provided in this application can be implemented in software. The video detection device can be stored in a memory and can be software in the form of programs and plug-ins. It includes a series of units, including an acquisition unit, a processing unit, and a determination unit. The acquisition unit, processing unit, and determination unit are used to implement the video detection method provided in this application.

[0109] In other feasible embodiments, the video detection device provided in this application embodiment can also be implemented in a combination of hardware and software. As an example, the video detection device provided in this application embodiment can be a processor in the form of a hardware decoding processor, which is programmed to execute the video detection method provided in this application embodiment. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0110] This application's embodiments determine the spoiler detection results of the recommended videos based on the video segment features and viewing status of each video segment of the target video and the video segment features of each video segment of the recommended videos to be detected. On the one hand, spoiler detection can be completed solely based on the video features (i.e., the features of video screen information). Compared to methods that combine video screen information, video subtitle information, and video bullet screen information for spoiler detection, the processing method is simpler and requires less data, thus enabling rapid spoiler detection. On the other hand, dividing the video into multiple video segments and performing spoiler detection based on the video segment features of each video segment allows for fine-grained feature comparison detection, thereby effectively improving the accuracy of the spoiler detection results. Based on the accurate spoiler detection results of the recommended videos to be detected, when displaying recommended videos associated with the target video to the user, spoiler videos of the target video can be accurately filtered out, and only non-spoiler videos of the target video can be displayed to the user. This effectively avoids the problem that users' interest in watching the original work is reduced due to watching spoiler videos, thus protecting the rights of the original target video, such as viewership.

[0111] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. The computer device described in this embodiment includes a processor 801, a communication interface 802, and a memory 803. The processor 801, communication interface 802, and memory 803 can be connected via a bus or other means; this embodiment uses a bus connection as an example.

[0112] The processor 801 (or CPU, Central Processing Unit) is the computing and control core of the computer device. It can parse various instructions and process various data within the computer device. For example, the CPU can parse power-on / off commands sent by the user and control the computer device to perform power-on / off operations; it can also transmit various interactive data between internal structures of the computer device. The communication interface 802 may optionally include standard wired interfaces or wireless interfaces (such as Wi-Fi, mobile communication interfaces, etc.), and is controlled by the processor 801 for sending and receiving data. The memory 803 is the storage device in the computer device, used to store programs and data. It is understood that the memory 803 here can include the computer device's built-in memory, or it can include extended memory supported by the computer device. The memory 803 provides storage space for the computer device's operating system, which may include, but is not limited to, Android, iOS, Windows Phone, etc., and this application does not limit this.

[0113] In this embodiment of the application, the processor 801 performs the following operations by running the executable program code in the memory 803:

[0114] Obtain a video feature set of the target video; wherein, the video feature set includes M target video features and the viewing state corresponding to each target video feature, each target video feature is a video feature of one of the M target video segments, the M target video segments are obtained by video segmentation of the target video; different target video features correspond to different target video segments, and the viewing state corresponding to a target video feature is the viewing state of the target video segment corresponding to the target video feature; M is a positive integer greater than 1;

[0115] Obtain recommended videos to be detected from the video recommendation pool, and perform video segmentation on the recommended videos to be detected to obtain N recommended video segments, where N is a positive integer greater than 1;

[0116] The video features of each of the recommended video segments are determined to obtain N recommended video features; based on the video feature set of the target video and the N recommended video features, the spoiler detection result of the recommended video to be detected is determined.

[0117] In one possible implementation, the processor 801 is further configured to:

[0118] Obtain the target video, and segment the target video to obtain M target video segments;

[0119] The video features of each target video segment are determined to obtain M target video features; the viewing state of each target video segment is determined, and based on the viewing state of each target video segment, the correspondence between the M target video segments and the M target video features, the viewing state corresponding to each target video feature is determined; a subset of viewed video features is generated based on the target video features whose corresponding viewing state is "viewed", and a subset of unviewed video features is generated based on the target video features whose corresponding viewing state is "not viewed". The video feature set of the target video is generated based on the subset of viewed video features and the subset of unviewed video features.

[0120] In one possible implementation, when the processor 801 determines the spoiler detection result of the recommended video to be detected based on the video feature set of the target video and the N recommended video features, it is specifically used for:

[0121] The N recommended video features are compared with each target video feature in the subset of watched video features, and the N recommended video features are compared with each target video feature in the subset of unwatched video features to obtain comparison results.

[0122] If, based on the comparison results, it is determined that a first type of recommended video feature exists among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined as the first spoiler detection result.

[0123] If, based on the comparison results, it is determined that the first type of recommended video feature does not exist among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined to be the second spoiler detection result.

[0124] Wherein, the first spoiler detection result indicates that the recommended video to be detected is a spoiler video of the target video, and the second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video; the first type of recommended video feature has a matching target video feature in the unwatched video feature subset, but no matching target video feature in the watched video feature subset.

[0125] In one possible implementation, the processor 801 is further configured to:

[0126] If, based on the comparison results, K of the N recommended video features are determined to be second-type recommended video features, and NK of the remaining N recommended video features are third-type recommended video features, then it is determined that the first-type recommended video features do not exist among the N recommended video features.

[0127] Wherein, K is an integer greater than or equal to 1 and less than or equal to N; the second type of recommended video feature has a matching target video feature in the subset of watched video features; the third type of recommended video feature does not have a matching target video feature in the subset of watched video features, nor does it have a matching target video feature in the subset of unwatched video features.

[0128] In one possible implementation, the processor 801 is further configured to:

[0129] Determine the playback time corresponding to each of the first type of recommended video features, wherein the playback time is the segment playback time of the target video segment corresponding to the target video feature that matches the first type of recommended video feature in the subset of unwatched video features;

[0130] Obtain new viewing time information for the target video, including the playback time of the already viewed portion of the target video;

[0131] If the playback time of the viewed portion of the video includes the playback time corresponding to each of the first type of recommended video features, then the spoiler detection result of the recommended video to be detected is adjusted from the first spoiler detection result to the second spoiler detection result, and the second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video.

[0132] In one possible implementation, when the processor 801 performs video segmentation on the recommended video to be detected to obtain N recommended video segments, it is specifically used for:

[0133] Video segmentation information is determined from the recommended videos to be detected. The video segmentation information includes one or both of black frame information and scene switching frame information.

[0134] Based on the video segmentation information, the recommended video to be detected is segmented to obtain N recommended video segments.

[0135] In one possible implementation, the processor 801 is further configured to place the recommended video to be detected into the spoiler video pool of the target video if the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a spoiler video of the target video;

[0136] If the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a non-spoiler video of the target video, then the recommended video to be detected is placed into the non-spoiler video pool of the target video.

[0137] In one possible implementation, the processor 801 is further configured to obtain target recommended videos from a pool of non-spoiler videos of the target video;

[0138] The relevant information of the target recommended video is sent to the client so that the client can display the relevant information of the target-detected recommended video on the video recommendation page corresponding to the target video in the video playback application.

[0139] This application's embodiments determine the spoiler detection results of the recommended videos based on the video segment features and viewing status of each video segment of the target video and the video segment features of each video segment of the recommended videos to be detected. On the one hand, spoiler detection can be completed solely based on the video features (i.e., the features of video screen information). Compared to methods that combine video screen information, video subtitle information, and video bullet screen information for spoiler detection, the processing method is simpler and requires less data, thus enabling rapid spoiler detection. On the other hand, dividing the video into multiple video segments and performing spoiler detection based on the video segment features of each video segment allows for fine-grained feature comparison detection, thereby effectively improving the accuracy of the spoiler detection results. Based on the accurate spoiler detection results of the recommended videos to be detected, when displaying recommended videos associated with the target video to the user, spoiler videos of the target video can be accurately filtered out, and only non-spoiler videos of the target video can be displayed to the user. This effectively avoids the problem that users' interest in watching the original work is reduced due to watching spoiler videos, thus protecting the rights of the original target video, such as viewership.

[0140] This application also provides a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to execute the video detection method described in this application. The specific implementation can be found in the foregoing description and will not be repeated here.

[0141] This application also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the video detection method as described in this application. Specific implementation details are provided above and will not be repeated here.

[0142] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0143] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0144] The above-disclosed embodiments are only some of the embodiments of this application, and should not be construed as limiting the scope of this application. Therefore, any equivalent changes made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A video detection method, characterized in that, The method includes: Obtain a video feature set of the target video; wherein, the video feature set includes: a subset of viewed video features and a subset of unviewed video features; the subset of viewed video features is generated based on target video features with a viewed state among M target video features, and the subset of unviewed video features is generated based on target video features with an unviewed state among the M target video features; each target video feature is a video feature of one of the M target video segments, and the M target video segments are obtained by video segmentation of the target video; different target video features correspond to different target video segments, and the viewing state corresponding to a target video feature is the viewing state of the target video segment corresponding to the target video feature; M is a positive integer greater than 1; Obtain recommended videos to be detected from the video recommendation pool, and perform video segmentation on the recommended videos to be detected to obtain N recommended video segments, where N is a positive integer greater than 1; Determine the video features of each of the recommended video segments to obtain N recommended video features; The N recommended video features are compared with each target video feature in the subset of watched video features, and the N recommended video features are compared with each target video feature in the subset of unwatched video features to obtain comparison results. If, based on the comparison results, it is determined that a first type of recommended video feature exists among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined as the first spoiler detection result. If, based on the comparison results, it is determined that the first type of recommended video feature does not exist among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined to be the second spoiler detection result. Wherein, the first spoiler detection result indicates that the recommended video to be detected is a spoiler video of the target video, and the second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video; the first type of recommended video feature has a matching target video feature in the unwatched video feature subset, but no matching target video feature in the watched video feature subset.

2. The method according to claim 1, characterized in that, The method further includes: Obtain the target video, and segment the target video to obtain M target video segments; Determine the video features of each target video segment to obtain M target video features; Determine the viewing status of each target video segment, and based on the viewing status of each target video segment, the correspondence between the M target video segments and the M target video features, determine the viewing status corresponding to each target video feature; A subset of viewed video features is generated based on the features of the target video whose corresponding viewing state is "viewed". A subset of unviewed video features is generated based on the features of the target video whose corresponding viewing state is "not viewed". A video feature set of the target video is generated based on the subset of viewed video features and the subset of unviewed video features.

3. The method according to claim 1, characterized in that, The method further includes: If, based on the comparison results, K of the N recommended video features are determined to be second-type recommended video features, and NK of the remaining N recommended video features are third-type recommended video features, then it is determined that the first-type recommended video features do not exist among the N recommended video features. Wherein, K is an integer greater than or equal to 1 and less than or equal to N; the second type of recommended video feature has a matching target video feature in the subset of watched video features; the third type of recommended video feature does not have a matching target video feature in the subset of watched video features, nor does it have a matching target video feature in the subset of unwatched video features.

4. The method according to claim 1, characterized in that, The method further includes: Determine the playback time corresponding to each of the first type of recommended video features, wherein the playback time is the segment playback time of the target video segment corresponding to the target video feature that matches the first type of recommended video feature in the subset of unwatched video features; Obtain new viewing time information for the target video, including the playback time of the already viewed portion of the target video; If the playback time of the viewed portion of the video includes the playback time corresponding to each of the first type of recommended video features, then the spoiler detection result of the recommended video to be detected is adjusted from the first spoiler detection result to the second spoiler detection result, and the second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video.

5. The method according to any one of claims 1-4, characterized in that, The process involves segmenting the recommended video to be detected into N recommended video segments, including: Video segmentation information is determined from the recommended videos to be detected. The video segmentation information includes one or both of black frame information and scene switching frame information. Based on the video segmentation information, the recommended video to be detected is segmented to obtain N recommended video segments.

6. The method according to any one of claims 1-4, characterized in that, The method further includes: If the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a spoiler video of the target video, then the recommended video to be detected is placed into the spoiler video pool of the target video; If the spoiler detection result of the recommended video to be detected indicates that the recommended video to be detected is a non-spoiler video of the target video, then the recommended video to be detected is placed into the non-spoiler video pool of the target video.

7. The method according to claim 6, characterized in that, The method further includes: Obtain target recommended videos from the non-spoiler video pool of the target video; The relevant information of the target recommended video is sent to the client so that the client can display the relevant information of the target recommended video on the video recommendation page corresponding to the target video in the video playback application.

8. A video detection device, characterized in that, The device includes: An acquisition unit is used to acquire a video feature set of a target video; wherein, the video feature set includes: a subset of viewed video features and a subset of unviewed video features; the subset of viewed video features is generated based on target video features with a viewed state among M target video features, and the subset of unviewed video features is generated based on target video features with an unviewed state among the M target video features; each target video feature is a video feature of one of the M target video segments, and the M target video segments are obtained by video segmentation of the target video; different target video features correspond to different target video segments, and the viewing state corresponding to a target video feature is the viewing state of the target video segment corresponding to the target video feature; M is a positive integer greater than 1; The processing unit is used to obtain the recommended video to be detected from the video recommendation pool, and to perform video segmentation on the recommended video to be detected to obtain N recommended video segments, where N is a positive integer greater than 1; A determining unit is used to determine the video features of each of the recommended video segments to obtain N recommended video features; The processing unit is further configured to: compare the N recommended video features with each target video feature in the subset of watched video features, and compare the N recommended video features with each target video feature in the subset of unwatched video features to obtain a comparison result; if the comparison result determines that a first type of recommended video feature exists among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined to be a first spoiler detection result; if the comparison result determines that the first type of recommended video feature does not exist among the N recommended video features, then the spoiler detection result of the recommended video to be detected is determined to be a second spoiler detection result. Wherein, the first spoiler detection result indicates that the recommended video to be detected is a spoiler video of the target video, and the second spoiler detection result indicates that the recommended video to be detected is a non-spoiler video of the target video; the first type of recommended video feature has a matching target video feature in the unwatched video feature subset, but no matching target video feature in the watched video feature subset.

9. A computer device, characterized in that, The computer device includes an input interface and an output interface, and the computer device further includes: A processor, suitable for implementing one or more computer programs; and, A computer-readable storage medium storing one or more computer programs adapted to be loaded by the processor and executed as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more computer programs adapted to be loaded by a processor and executed as described in any one of claims 1-7.

11. A computer program product, characterized in that, The computer program product includes a computer program stored in a computer-readable storage medium, a processor of a computer device reads the computer program from the computer-readable storage medium, and the processor executes the computer program, causing the computer device to perform the video detection method as described in any one of claims 1-7.

Images